Machine-Readable Endpoints

fiskaly publishes a set of machine-readable endpoints alongside the human-readable documentation site. These endpoints are designed for AI agents, LLM crawlers, and automation pipelines that need structured access to product metadata, API specifications, and regulatory data.

Endpoint reference

Endpoint	Format	Purpose
/llms.txt	Plain text	LLM discovery index — lists all documentation pages and spec URLs
/llms-full.txt	Plain text	Full documentation dump for LLM context ingestion
/products.json	JSON	Product manifest with metadata, base URLs, and spec URLs for all 14 products
/human-interventions.json	JSON	Registry of automatable vs human-required actions with summary counts
/regulatory.json	JSON	Per-country regulatory requirements and compliance details
/.well-known/ai-plugin.json	JSON	AI plugin manifest (OpenAI plugin spec format)
/CLAUDE.md	Markdown	Agent integration guide with authentication, patterns, and per-country reference
/specs/sign-de-v2.json	JSON	SIGN DE v2 OpenAPI specification
/specs/spec-sign-at.json	JSON	SIGN AT (RKSV) OpenAPI specification
/specs/fiskaly.sign-fr.2026-02-03.yaml	YAML	SIGN FR OpenAPI specification (Unified API)
/specs/fiskaly.sign-it.2026-02-03.yaml	YAML	SIGN IT OpenAPI specification (Unified API)
/specs/fiskaly.e-invoice-be.2026-02-03.yaml	YAML	E-INVOICE BE OpenAPI specification (Unified API)
/specs/fiskaly.e-invoice-de.2026-02-03.yaml	YAML	E-INVOICE DE OpenAPI specification (Unified API)
/specs/spec-sign-es.yaml	YAML	SIGN ES OpenAPI specification
/specs/spec-dsfinvk.json	JSON	DSFinV-K OpenAPI specification
/specs/spec-ereceipt.json	JSON	eReceipt OpenAPI specification
/specs/spec-merchant.json	JSON	Management API OpenAPI specification
/specs/spec-obono.json	JSON	obono (SIGN AT SaaS) OpenAPI specification

All endpoints are served from https://workspace.fiskaly.com.

Discovery flow

A typical agent discovery flow looks like this:

1. Fetch /llms.txt
   → Get an index of all available documentation and specs

2. Fetch /products.json
   → Find the product matching the target country
   → Extract baseUrls, specUrl, and docsUrl

3. Fetch /specs/{product-spec}
   → Load the full OpenAPI spec for the chosen product

4. Fetch /human-interventions.json
   → Identify which steps are automatable vs human-required

products.json

The product manifest is the primary structured data endpoint. Each product entry includes:

{
  "id": "sign-de",
  "name": "SIGN DE",
  "description": "German fiscal signing (KassenSichV)",
  "country": "DE",
  "apiArchitecture": "specialized",
  "baseUrls": {
    "test": "https://kassensichv-middleware.fiskaly.com/api/v2",
    "live": "https://kassensichv.fiskaly.com/api/v2"
  },
  "specUrl": "/specs/sign-de-v2.json",
  "docsUrl": "/countries/germany",
  "apiReferenceUrl": "/api/sign-de"
}

Use apiArchitecture to determine the integration pattern:

• specialized — per-country base URL, client-generated UUIDs, separate Management API
• unified — shared base URLs, server-generated IDs, required headers (X-Api-Version, X-Idempotency-Key, X-Scope-Identifier)
• platform — cross-product services (DSFinV-K, eReceipt, Management API)

human-interventions.json

Returns all integration actions categorized by automation level:

{
  "summary": {
    "fullyAutomatable": 11,
    "partiallyAutomatable": 2,
    "requiresHuman": 7
  },
  "interventions": [
    {
      "action": "Authenticate (get token)",
      "product": "All",
      "automatable": "yes",
      "reason": "Standard API call",
      "whatToDo": "POST /auth with api_key + api_secret"
    }
  ]
}

See the Human Intervention Registry for the full human-readable reference.

ai-plugin.json

The AI plugin manifest follows the OpenAI plugin specification format:

{
  "schema_version": "v1",
  "name_for_model": "fiskaly",
  "description_for_model": "fiskaly provides fiscal compliance APIs...",
  "api": {
    "type": "openapi",
    "url": "https://workspace.fiskaly.com/specs/sign-de-v2.json"
  }
}

Crawler access

The robots.txt configuration allows LLM crawlers to access all documentation and machine-readable endpoints. Standard web crawlers are permitted for documentation pages, spec files, and structured data endpoints.

💡Tip

Prefer the MCP server over raw HTTP fetching when working in Claude Code or Cursor. The MCP server provides the same data through structured tool calls with search and filtering built in. See the MCP Server guide.